KOC-WebPredictor: Soil Sorption Coefficient Predictions for 7,612 Chemicals

An interpretable QSAR and machine learning model for predicting soil organic carbon–water partition coefficients (log KOC), a key determinant of contaminant mobility. The model was developed on 824 structurally diverse compounds and used to screen 7,612 chemicals from the U.S. EPA's CPDat inventory. The dataset, authored by Lu Li and last updated in March 2026, includes an open-access web tool for predictions.

Use Cases

• Predict soil sorption coefficients for new chemicals based on molecular descriptors like hydrophobicity and aromatic rigidity.
• Screen large chemical inventories for environmental mobility potential based on predicted log KOC values.
• Prioritize chemicals for experimental testing under regulatory frameworks using the model's applicability domain.
• Gain mechanistic insights into soil–organic carbon interactions from the interpretable QSAR model.

Strengths

• Model built on 824 structurally diverse compounds with strong reported performance (R² = 0.85).
• Large-scale external screening of 7,612 chemicals with 94% coverage within the model's applicability domain.
• Includes an open-access web tool for quantitative and qualitative predictions with visualization.

Limitations

• Row count for the underlying dataset is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

figshare, authored by Lu Li.

Collection Method

Developed using QSAR and machine learning approaches on experimental KOC data and molecular descriptors.

Time Range

Model development date not specified; dataset last updated March 2026.

Freshness

Last updated 2026-03-17 13:06:28; freshness should be verified.

Geography

Chemical inventory sourced from U.S. EPA's CPDat; geographic scope of underlying data not specified.